Detailed Case Study

Background & Context

Demand forecasting directly feeds wave management and inbound planning—core EWM functions. Supply/demand reconciliation logic demonstrates understanding of business operations, not just data science. This project shows how analytics translates into operational decisions.

Traditional SAP EWM wave creation relies on current order volume and static rules. This project adds predictive capability, allowing planners to anticipate future demand and optimize wave scheduling proactively.

Challenge Analysis

Key challenges addressed:

• SKU-Level Forecasting: Need to forecast demand for individual SKUs, not just aggregate
• Wave Planning Integration: Forecasts must translate into actionable wave planning suggestions
• Seasonality Handling: Weekly patterns, monthly trends, and holiday effects
• SAP Integration: Forecast outputs must be consumable by SAP EWM systems
• Operational Relevance: Not just a forecast, but aligned with warehouse operations

Solution Architecture

The system follows a per-SKU forecasting approach:

┌─────────────────────┐
│ Historical Demand   │
│ (Daily, per SKU)    │
│ 6+ months history   │
└──────────┬──────────┘
           │
           ▼
┌─────────────────────┐
│ Prophet Model       │
│ Per SKU             │
│ - Trend             │
│ - Weekly Seasonality│
│ - Holiday Effects   │
└──────────┬──────────┘
           │
           ▼
┌─────────────────────┐
│ 4-Week Forecast     │
│ + Confidence        │
│   Intervals         │
└──────────┬──────────┘
           │
           ▼
┌─────────────────────┐
│ Wave Planning       │
│ Suggestions         │
│ - Expected lines    │
│ - Peak day flags    │
│ - Wave windows      │
└─────────────────────┘

Key Components:

1. Data Preparation: Aggregates daily demand by SKU from EWM transaction data
2. Forecasting Engine: Prophet model per SKU with seasonality and holidays
3. Wave Planning Logic: Simple heuristics to suggest wave windows
4. OData Service: Exposes forecasts for SAP EWM consumption

Implementation Details

Data Requirements

Single warehouse, top 20 SKUs with:

• Minimum 6 months of daily history
• Date, SKU, demand quantity fields
• Optional holiday/promotion calendar

Forecasting Approach

Prophet was chosen for its strengths:

• Handles seasonality and trends well
• Built-in holiday support
• Provides confidence intervals
• Easy to interpret and explain to stakeholders

Code Example: Prophet Forecasting

from prophet import Prophet
import pandas as pd

def forecast_sku_demand(sku_data):
    """Forecast demand for a single SKU."""
    # Prepare data
    df = sku_data[['date', 'demand_qty']].rename(
        columns={'date': 'ds', 'demand_qty': 'y'}
    )
    
    # Initialize model with seasonality
    model = Prophet(
        yearly_seasonality=False,
        weekly_seasonality=True,
        daily_seasonality=False
    )
    
    # Add holidays if available
    if 'holidays' in sku_data.columns:
        model.add_country_holidays(country_name='US')
    
    # Fit and forecast
    model.fit(df)
    future = model.make_future_dataframe(periods=28)  # 4 weeks
    forecast = model.predict(future)
    
    return forecast

Wave Planning Logic

Simple heuristic-based wave suggestions:

• Calculate expected lines to pick from forecasted demand
• Identify peak days (exceeding 1.5x average)
• Suggest wave windows: morning, afternoon, evening
• Recommend 2-3 waves based on volume thresholds

Results & Metrics

The system delivers operational value:

Lessons Learned

• Operational Alignment: Forecasts must connect to warehouse operations, not just be accurate
• SQL Skills Transfer: SQL proficiency transfers to SAP Open SQL for extracting EWM transaction data
• OData Integration: Understanding of modern SAP integration patterns
• CDS Views: Demonstrates S/4HANA cloud-ready architecture thinking